Working with 3D meshes

2021-02-11

library(facefuns)
library(Morpho)
library(rgl)

WORK IN PROGRESS

This tutorial covers basic steps in working with 3D meshes. There are multiple ways to get your 3D data into R - the workflow outlined here is just one of many options!

Importing 3D data into R

To conduct a GPA and shape PCA, you’ll need your shape data in a three-dimensional array, just as for 2D data. In the Face Research Lab, we’re using a DI3D system (you can find more info on our specific setup here). 3D meshes are exported as OBJs and extraneous parts of the meshes manually removed (using MeshLab), before meshes are delineated and warped to a standard face in MorphAnalyser¹.

OBJ vs PLY

OBJs can be read directly into R using rgl::readOBJ. readOBJ imports the OBJ as an object of class “mesh3d”. Our OBJs and resulting mesh objects are fairly big (~10MB), which makes them a wee bit unwieldy.

Instead of working with OBJs, you can also convert them into smaller PLY files. PLYs come in different flavours: ASCII, binary_little_endian or binary_big_endian. geomorph has a dedicated read.ply function, which reads PLY files of the ASCII type. If you only have a couple of files, you could use MeshLab to manually convert your OBJs into ASCII PLYs.

If you have a bigger number of files, you might be looking for a batch solution. I’m using MeshLab’s command line interface (MeshLabServer) to do so. If you’ve not much used the command line before, don’t be put off - while setting everything up can be a bit tedious initially, the actual running of commands is very little fuss. [write extra tutorial on that?]

The only catch here is that other than the MeshLab GUI, MeshLabServer can only export binary PLY files (or at least that was the case in 2018). Instead of using geomorph, we’ll thus be using the package Morpho to import our data.

Read PLYs

First, we’ll specify the path to our data

path_to_ply <- system.file("extdata", "ply", package="facefuns")
  
path_to_example <- list.files(path = path_to_ply,
                          pattern = "\\.ply$",
                          full.names = TRUE)[1]

Now, we can use Morpho::file2mesh to import the first of the 3D example faces included in facefuns

plyexample <- Morpho::file2mesh(path_to_example)

You can use the package rgl to plot the face

open3d()
#> glX 
#>   1
plot3d(plyexample,
       type = "wire", col = "gray",
       box = FALSE, axes = FALSE,
       xlab = "", ylab = "", zlab = "",
       specular = "black")
aspect3d("iso")
rglwidget()

close3d()

As you can see in the environment, this mesh3d object is still very large (6.7MB). All we really need at this point are the vertex coordinates. facefuns::read_ply reads PLYs using Morpho::file2mesh, extracts only the vertex data and saves it into an array that we can then use for subsequent analyses.

plydata <- read_ply(path_to_ply)
str(plydata)
#>  num [1:60541, 1:3, 1:12] -77.4 -77.4 -77.3 -77.3 -77.2 ...
#>  - attr(*, "dimnames")=List of 3
#>   ..$ : NULL
#>   ..$ : NULL
#>   ..$ : chr [1:12] "avg_f18" "avg_f26" "avg_f36" "avg_f46" ...

References

1. Tiddeman, B. P., Duffy, N., & Rabey, G. (2000). Construction and visualisation of three-dimensional facial statistics. Computer Methods and Programs in Biomedicine, 63, 9–20. https://doi.org/10.1016/S0169-2607(00)00072-9